Landslide risk assessment and management: an overview
F.C. Dai a
, C.F. Lee b,*, Y.Y. Ngai b
a
Institute of Geographical Sciences and Natural Resources, Chinese Academy of Sciences, Beijing 100101, China
b
Department of Civil and Structural Engineering, The University of Hong Kong, Pokfulam Road, Hong Kong, China
Received 21 December 2000; accepted 29 August 2001
Abstract
Landslides can result in enormous casualties and huge economic losses in mountainous regions. In order to mitigate
landslide hazard effectively, new methodologies are required to develop a better understanding of landslide hazard and to make
rational decisions on the allocation of funds for management of landslide risk. Recent advances in risk analysis and risk
assessment are beginning to provide systematic and rigorous processes to enhance slope management. In recent years, risk
analysis and assessment has become an important tool in addressing uncertainty inherent in landslide hazards. This article
reviews recent advances in landslide risk assessment and management, and discusses the applicability of a variety of approaches
to assessing landslide risk. Firstly, a framework for landslide risk assessment and management by which landslide risk can be
reduced is proposed. This is followed by a critical review of the current state of research on assessing the probability of
landsliding, runout behavior, and vulnerability. Effective management strategies for reducing economic and social losses due to
landslides are described. Problems in landslide risk assessment and management are also examined. D 2002 Elsevier Science
B.V. All rights reserved.
Keywords: Landslides; Probability; Runout behavior; Vulnerability; Risk
1. Introduction
Landslides, as one of the major natural hazards,
account each year for enormous property damage in
terms of both direct and indirect costs. Landslides,
defined as the movement of a mass of rock, debris or
earth down a slope (Cruden, 1991), can be triggered
by a variety of external stimulus, such as intense
rainfall, earthquake shaking, water level change,
storm waves or rapid stream erosion that cause a
rapid increase in shear stress or decrease in shear
strength of slope-forming materials. In addition, as
development expands into unstable hillslope areas
under the pressures of increasing population and
urbanization, human activities such as deforestation
or excavation of slopes for road cuts and building
sites, etc., have become important triggers for landslide
occurrence.
Landslides have caused large numbers of casualties
and huge economic losses in mountainous areas
of the world. The most disastrous landslides have
claimed as many as 100,000 lives (Li and Wang,
1992). In the United States, landslides cause an
estimated US$1–2 billion in economic losses and
0013-7952/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
PII: S0013-7952(01)00093-X
*
Corresponding author. Tel.: +852-285-92-645; fax: +852-285-
80-611.
E-mail address: leecf@hkucc.hku.hk (C.F. Lee).
www.elsevier.com/locate/enggeo
Engineering Geology 64 (2002) 65–87
about 25–50 deaths annually, thus exceeding the
average losses due to earthquakes (Schuster and
Fleming, 1986). Li and Wang (1992) conservatively
estimated that in China the number of deaths caused
by landslides totaled more than 5000 during the
1951–1989 period, resulting in an average of more
than 125 deaths annually, and annual economic losses
of about US$500 million.
Social and economic losses due to landslides can
be reduced by means of effective planning and
management. These approaches include: (a) restriction
of development in landslide-prone areas, (b) use
of excavation, grading, landscaping, and construction
codes, (c) use of physical measures (drainage,
slope-geometry modification, and structures) to prevent
or control landslides, and (d) development of
warning systems (Slosson and Krohn, 1982; Schuster
and Leighton, 1988; Schuster, 1996). Schuster
and Leighton (1988) estimated that these methods
could reduce landslide losses in California by more
than 90%. Slosson and Krohn (1982) stated that
enactment of these approaches had already reduced
landslide losses in the City of Los Angeles by 92–
97%. However, in spite of improvements in hazard
recognition, prediction, mitigation measures, and
warning systems, worldwide landslide activity is
increasing. This trend is expected to continue in
the 21st century for the following reasons (Schuster,
1996):
 increased urbanization and development in
landslide-prone areas;
 continued deforestation of landslide-prone
areas; and
 increased regional precipitation caused by
changing climatic patterns.
To address the landslide problem, governmental
agencies need to develop a better understanding of
landslide hazard and to make rational decisions on
allocation of funds for management of landslide risk.
However, it is widely accepted that the landslide
problem is dominated by uncertainty. This uncertainty
arises at all stages in the resolution of the
problem, from site characterization to material property
evaluation to analysis and design and consequence
assessment (Morgenstern, 1997). Recent
advances in risk analysis and risk assessment are
beginning to provide systematic and rigorous processes
to formalize slope engineering practice and
enhance slope management (Fell and Hartford,
1997). In recent years, risk analysis and assessment
has become an important tool in addressing uncertainty
inherent in landslide hazards.
This article reviews recent advances in landslide
risk assessment and management, and discusses the
applicability of a variety of approaches to assessing
landslide risk. Since various definitions of risk terms
are available, the authors have adopted the definitions
used in the IUGS Working Group on Landslides,
Committee on Risk Assessment (1997).
2. Basic framework for landslide risk assessment
and management
Landslide risk assessment and management comprises
the estimation of the level of risk, deciding
whether or not it is acceptable, and exercising appropriate
control measures to reduce the risk when the
risk level cannot be accepted (Ho et al., 2000). It
requires the following issues to be addressed: (a)
probability of landsliding, (b) runout behavior of
landslide debris, (c) vulnerability of property and
people to landslide, (d) landslide risk to property
and people, and (e) management strategies and decision-making
(Fig. 1).
Fig. 1. A framework for landslide risk assessment and management.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8766
In terms of conditional probability, landslide risk
when defined as the annual probability of loss of life
of a specific individual may be calculated as follows
(Morgan et al., 1992):
RðDIÞ ¼ PðHÞ Â PðS j HÞ Â PðT j SÞ
 VðL j TÞ ð1Þ
where R(DI) is the risk (annual probability of loss of
life to an individual); P(H) is the annual probability of
the landslide event; P(SjH) is the probability of spatial
impact given the event; P(TjS) is the probability of
temporal impact given the spatial impact; and V(LjT)
is the vulnerability of the individual (probability of
loss of life of the individual given impact).
For a case involving property damage the equivalent
expression would be
RðPDÞ ¼ PðHÞ Â PðS j HÞ Â VðP j SÞ Â E ð2Þ
where R(PD) is the risk (annual loss of property
value); P(H) is the annual probability of the landslide
event; P(SjH) is the probability of spatial impact (i.e.
of the landslide impacting the property); V(PjS) is the
vulnerability of the property (proportion of property
value lost); E is the element at risk (e.g. the value of
the property).
Some researchers considered the product of
P(SjH) Â P(TjS) Â V(LjT) in Eq. (1) or P(SjH) Â V
(PjS) Â E in Eq. (2) as ‘‘consequence’’ (e.g. Wong
et al., 1997) or the product of P(H) Â P(SjH) as
‘‘hazard’’ (e.g. Leroueil and Locat, 1998).
Landslide risk can be assessed qualitatively or
quantitatively. Whether qualitative or quantitative
assessments are more suitable depends on both the
desired accuracy of the outcome and the nature of the
problem, and should be compatible with the quality
and quantity of available data. Generally, for a large
area where the quality and quantity of available data
are too meager for quantitative analysis, a qualitative
risk assessment may be more applicable; while for
site-specific slopes that are amenable to conventional
limit equilibrium analysis, a detailed quantitative risk
assessment should be carried out.
3. Assessment of probability of landsliding
When assessing the probability of landsliding
within a specified period of time and within a given
area, recognition of the conditions that caused the
slope to become unstable and the processes that
triggered the movement is of primary importance.
The factors which determine the probability of landsliding
for a particular slope or an area may be
grouped into two categories: (1) the preparatory
variables which make the slope susceptible to failure
without actually initiating it and thereby tending to
place the slope in a marginally stable state, such as
geology, slope gradient and aspect, elevation, soil
geotechnical properties, vegetation cover and longterm
drainage patterns and weathering; and (2) the
triggering variables which shift the slope from a
marginally stable to an unstable state and thereby
initiating failure in an area of given susceptibility,
such as heavy rainfall and earthquakes (Wu and Sidle,
1995). Obviously, the probability of landsliding depends
on both the preparatory and triggering variables.
However, the triggering variables may change
over a very short time span, and are thus very difficult
to estimate. If triggering variables are not taken into
account, the term ‘‘susceptibility’’ may be employed
to define the likelihood of occurrence of a landslide
event. At present, when assessing the probability of
landsliding on regional scales, it might be feasible to
consider landslide susceptibility as the probability of
landsliding based on the assumption that long-term
historic landslide records tend to smooth-out the
spatio-temporal effect of triggering factors on landslide
occurrence. For large-scale hazard assessments,
in which work is carried out in relatively small areas
or specific slopes, data collection at this scale should
relate to the quantitative parameters needed for slope
stability modeling.
Numerous methods have been developed to assess
the probability of landsliding. Soeters and van Westen
(1996) and van Westen et al. (1997) divided these
methods into inventory, heuristic, statistical, and deterministic
approaches.
The most straightforward initial approach to any
study of landslide hazard is the compilation of a landslide
inventory, and such inventories are the basis of
most susceptibility mapping techniques. On detailed
landslide inventory maps, the basic information for
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 67
evaluating and reducing landslide hazards or risk on a
regional or community level should be provided,
including the state of activity, certainty of identification,
dominant type of slope movement, primary
direction of movement, estimated thickness of material
involved in landsliding, and date(s) of known
activity for each landslide (Wieczorek, 1984). They
can be prepared by collecting historic information on
landslide events or from aerial photograph interpretation
coupled with field checking. The presentation of
landslides on inventory maps varies from detailed
features of the landslides to points representing locations
of the landslides, depending on both the nature
of problem and the size of failures. Inventory maps
can be used as an elementary form of hazard map
because they show the locations of recorded landslides.
Based on the landslide inventory map, landslide
density or landslide isopleth can be generated by
counting circles (Wright et al., 1974). The historic
frequency of landslides in an area can be determined
to provide realistic estimates of landslide probability
throughout a region where landslides have caused a
significant amount of damage. The trigger/landsliding
and frequency–magnitude relations that help understand
landslide probabilities may be derived from
landslide inventories. However, landslide inventory
and isopleth maps do not identify areas that may be
susceptible to landsliding unless landslides have already
occurred.
With the heuristic approaches, expert opinions are
used to estimate landslide potential from data on
preparatory variables. They are based on the assumption
that the relationships between landslide susceptibility
and the preparatory variables are known and
are specified in the models. A set of variables is then
entered into the model to estimate landslide susceptibility
(e.g. Gupta and Joshi, 1989). One problem
with the heuristic models is that they need long-term
information on the landslides and their causal factors
for the same site or for sites with similar geoenvironmental
conditions. However, the information
is, in most cases, not available. Other limitations of
this method are the reproducibility of the results and
the subjectivity of weightings and ratings of the
variables.
Deterministic approaches are based on slope stability
analyses, and are only applicable when the ground
conditions are fairly uniform across the study area and
the landslide types are known and relatively easy to
analyze. They have been widely used to assess landslide
probability in small areas (Terlien et al., 1995;
Wu and Sidle, 1995). For rainfall-induced failures,
these models couple shallow subsurface flow (i.e. the
pore pressure spatial distribution) caused by rainfalls
of various return periods, predicted soil thickness, and
landsliding of the soil mantle. For earthquake-induced
failures, a conventional seismic hazard analysis is
used to determine the peak ground accelerations
(PGA) for different return periods and the stability
of slopes when subjected to an earthquake with
various return periods is examined using a pseudostatic
analysis. Stability conditions are generally evaluated
by means of an infinite slope stability model,
where local equilibrium along a potential slip surface
is considered. The advantage of the deterministic
models is that they permit quantitative factors of
safety to be calculated with due consideration for
the variability of soil properties if necessary, while
the main problem is the high degree of simplification
that is usually necessary for the use of such models.
Another problem that limits the applicability of the
deterministic models is that data requirements for
deterministic models can be prohibitive, and frequently
it is impossible to acquire the input data
necessary to use the models effectively.
Statistical models involve the statistical determination
of the combinations of variables that have led to
landslide occurrence in the past. Statistical estimates
are made for areas currently free of landslides, but
where similar conditions exist. Conventional multivariate
statistical methods, such as multiple regression
analysis and discriminant analysis, have been used to
assess landslide susceptibility (e.g. Yin and Yan,
1988; Carrara et al., 1991). However, the use of
multivariate statistical models has always been hindered
by the need for continuous data. Categorical
data, such as geology, can be used but it commonly
involves the use of dummy variables to indicate the
presence or absence of a variable. This can result in an
enormous increase in the number of variables, with
the increase being directly related to the number of
categories in each explanatory variable. Moreover,
both techniques showed limited value when the
dependent variables takes only two values, that is,
whether landsliding occurs or not. Under these circumstances,
the assumption needed to test the hypothF.C.
Dai et al. / Engineering Geology 64 (2002) 65–8768
esis in regression analysis are violated. In such cases,
another multivariate technique, that of logistic regression
that is used for estimating the probability of an
event occurring, is applied (e.g. Carrara et al., 1991;
Chung and Fabbri, 1999). The advantage of logistic
regression modeling over other multivariate statistical
techniques including multiple regression analysis and
discriminant analysis is that the dependent variable
can have only two values — an event occurring or not
occurring, and that predicted values can be interpreted
as probability since they are constrained to fall in the
interval between 0 and 1. Statistical techniques are
generally considered the most appropriate approach
for landslide susceptibility mapping at medium scales
of 1:10,000–1:50,000, because on this scale it is
possible to map out in detail the occurrence of past
landslides, and to collect sufficient information on the
variables that are considered to be relevant to the
occurrence of landslides. However, one problem with
the use of multivariate statistical approaches in establishing
correlations between independent variables
and landslide susceptibility is that there is a potential
danger that such statistical methods, when used in a
black-box manner with inadequate consideration of
the mechanics of the physical processes involved, are
liable to result in very coarse and even misleading
regression correlations (Ho et al., 2000). This means
that they must be applied in a suitable mechanistic
framework.
For site-specific slopes, the probability of failure
is usually considered as simply the probability that
the factor of safety is less than unity. The performance
function of slopes, denoted by G(X ) where X is
the collection of random input parameters, is a
function which defines the failure or safety state of
a slope. The function is defined in such a way that
failure is implied when G(X ) < 0 and safety by
G(X ) > 0. The boundary defined by G(X ) = 0 separating
the safety and failure domains is called the
limit state boundary.
The performance function for a slope is usually
taken as one of the following formats
GðXÞ ¼ RðXÞ À SðXÞ ð3Þ
or
GðXÞ ¼ FðXÞ À 1 ð4Þ
where R(X ) is the resistance and S(X ) is the action,
and F(X ) is the factor of safety. The format of Eq. (4)
is more common because the safety of slopes is
traditionally characterized by the factor of safety.
The format of Eq. (3), however, is preferable because
it has a lower degree of non-linearity (Mostyn and
Fell, 1997). The performance function of a slope is
usually formulated using the simplified limit equilibrium
method, such as the ordinary method of slices,
simplified Bishop’s method, and simplified Janbu’s
method.
Once the performance function is defined, the
probability of failure of a slope can be estimated by
the following methods:
(1) The first-order-second-moment (FOSM) method.
This method characterizes the frequency distribution
of the factor of safety F in terms of its mean
value lF and standard deviation rF. The reliability
index b is may be computed from b =(lF À 1.0)/rF. b
may be regarded as an index of the degree of
uncertainty and it can be related to the probability of
failure if the frequency distribution is known.
(2) Monte Carlo simulation. This method involves
a computerized sampling procedure used to approximate
the probability distribution of the factor of
safety by repeating the analysis many times, especially
the target reliability to be evaluated is small. A
set of random numbers is generated for the random
variables according to the chosen frequency distributions
of the input parameters.
For some landslides where piezometric levels are
recorded over some period and rainfall data are
available, the relationship between piezometric levels
and rainfall can be modeled using physical or statistical
models (e.g. Fell et al., 1991). The probability
of piezometric level required for landsliding or reactivation
of a slide is then assessed by analyzing
rainfall for a given period. This method is ideal in
principle for site-specific, relatively deep-seated landslides.
However, in reality it is difficult to achieve any
accuracy in the modeling because of the complex
infiltration processes involved, heterogeneity of the
soil and rockmass in the slope, and groundwater
seeping into the slide from below and upslope area
and that seeping outward. It is also apparent that a
lengthy period of calibration is likely required, to
experience a range of rainfall and piezometric con-
ditions.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 69
4. Runout behavior of landslide debris
Delimiting the extent of endangered areas is fundamental
to landslide risk assessment. These require
accurate prediction of the runout behavior of a landslide,
such as how far and how fast a landslide travels
once mobilized. Generally, runout behavior is a set of
quantitative or qualitative spatially distributed parameters
that define the destructive potential of a landslide.
These parameters for the purpose of landslide
risk assessment mainly include (Wong et al., 1997;
Hungr et al., 1999):
 Runout distance—the distance from the landslide
source area to the distal toe of the deposition
area;
 Damage corridor width—the width of the area
subjected to landslide damage in the distal part
of the landslide path where impact on buildings
and other facilities occurs;
 Velocity—the velocity of travel within the
damage corridor which determines the potential
damage to facilities and the design parameters
of any required protective measures;
 Depth of the moving mass—which influences
the impact force of the landslide within the
damage corridor; and
 Depth of deposits—landslide deposits may
build up to a sufficient depth behind a structure
to cause its collapse.
4.1. Factors contributing to runout behavior of
landslide debris
A realistic estimate of runout behavior of a landslide
depends on an adequate understanding of the
generic factors that control travel. Relevant parameters
to consider include slope characteristics, mechanisms
of failure and modes of debris movement,
downhill path, and residual strength behavior of
sheared zones.
4.1.1. Slope characteristics
The important slope characteristics include slope
geometry, the nature of the slope-forming material,
and upslope influence zone. It is not difficult to
understand that slope geometry has an important
impact on runout behavior of landslide debris. The
motion behavior of the sliding material involves the
redistribution of the potential energy available at failure
into friction energy, disaggregating or remolding
energy, and kinetic energy. Leroueil et al. (1996) and
Leroueil and Locat (1998) examine the relationship
between the nature of the slope-forming material and
slope movements. The convergence of hydrologic
pathways along catchment provides water for incorporation
into an initial failure. If a landslide occurs
during a runoff-producing storm, then overland flow
down the catchment axis and throughflow emanating
from the headscarp will spill into an initial failure.
Deformation accompanying an initial failure may
allow further incorporation of water emanating from
bedrock springs and surface runoff into the failed
material, thus increasing debris mobility (Montgomery
et al., 1991).
4.1.2. Mechanisms of failure and modes of debris
movement
Certain failure mechanisms such as collapse of
loose soil leading to static liquefaction and large-scale
rock fall may release mobile debris (Wong et al.,
1997). Loose granular soils tend to collapse when
sheared, which under undrained condition results in
an increase in pore water pressure (Fleming et al.,
1989). Consequently, failures in contractive soils
often evolve into debris flows that may travel great
distances because even minor strain may cause liquefaction.
Dilatant soils, on the other hand, expand upon
shearing and a continued influx of water is required to
sustain their mobilization. Consequently, failures in
dilatant soils tend to be relatively slow-moving slides,
depending on the availability of rainfall amount and
intensity and soil permeability (Fleming et al., 1989;
Dai et al., 1999, 2000). Once a landslide mobilizes,
the modes of debris movement, disintegration of the
failure debris during motion and convergence of surface
runoff obviously influence debris velocity and
travel distance. The availability of water further
affects whether it assumes characteristics typical of
debris flows, hyperconcentrated flows, or sedimentladen
floodwaters.
4.1.3. Downhill path
The characteristics of the downhill path traversed
by the debris can affect the mode of debris travel.
Important parameters include the gradient of the
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8770
downslope path, possibility of channelization of debris,
characteristics of ground surface on which the
debris travels, e.g. susceptibility to depletion, response
to rapid loading as a result of sudden debris
impact, type of vegetation, extent of catchment which
collects surface water and discharges into the downslope
area, etc. The topography downslope of an
initial failure affects the probability of debris mobilization.
An increase in downslope gradient will favor
the acceleration of an initial failure, whereas gentler
slopes will tend to inhabit acceleration. The studies
carried out by Nicoletti and Sorriso-Valvo (1991)
indicated that the downhill morphology of the slopes
and valleys where landslide debris moved exerts a
great influence on the damage corridor width and
runout distance of rock avalanches. The amount of
water available for mixing with landslide debris and
the gradient of the downslope channelway contribute
to the transition of an initial landslide into a mobile
debris flow. Incorporation of excessive volumes of
water may dilute landslide debris and increase its
mobility.
4.1.4. Residual strength behavior of sheared zones
The presence or absence of pre-existing shears
and the degree of brittleness is particularly important
(Hutchinson, 1995). If the material displays
strain-softening behavior, kinetic energy can reach
catastrophic proportions and long runout distance
can occur (Leroueil et al., 1996). Brittleness on preexisting
shears is generally low or zero. Hence,
reactivation of movement on such shears is usually
slow. The post-failure movement of landslides is
generally controlled by the residual strength behavior
of sheared zones. In this regard, knowledge of
the effect of rate of displacement on the residual
strength is necessary when studying the kinematics
of a potential sliding mass. Three types of variation
of residual strength with an increasing rate of
displacement have been recognized (Tika et al.,
1996): (a) neutral rate effect — soils showing a
constant residual strength irrespective of the rate
of displacement; (b) negative rate effect — soils
showing a significant drop in strength when sheared
at rates higher than a critical value; and (c) positive
rate effect— soils showing an increase in residual
strength above the slowly drained residual value at
increasing rates of displacement. This is of great
significance in predicting the velocity and displacement
of landslides. In a first-time landslide induced
in brittle soil, the loss in strength from the peak to
the residual value is likely to exceed the reduction
in shear stress due to changing geometry. Whatever
causing the initial failure, the landslide ceases to be
in equilibrium and moves to a new gentler position.
If the rate effect is positive, the strength increases
with velocity and the landslide decelerates. The
landslide moves slowly, with small momentum,
and comes to rest when it is in equilibrium with
its residual strength. If the rate effect on the residual
strength is negative, the landslide accelerates and
develops into a fast movement. In this case, long
runout distance may occur. For a reactivated landslide,
it is in limit equilibrium with the constant
strength, since the slow residual strength does not
change with displacement. In a soil with a positive
rate effect, extra strength is available to resist
movement at all displacements and displacement
rates, and the landslide thus decelerates and comes
to a rest. If the soil has a negative rate effect and if
during instability a critical combination of displacement
and velocity is also induced, the landslide
accelerates to potentially catastrophic velocities,
and large displacements occur. For instance, Tika
and Hutchinson (1999) explained or at least partially
explained why the Vaiont landslide developed into a
catastrophic disaster from the negative rate effect of
soil from the slip surface.
4.2. Methods for predicting runout distance of
landslide debris
Current and past research into the runout behavior
of a landslide can generally be grouped into three
categories. The first includes empirical models aimed
at providing practical tools for predicting the runout
distance and distribution of landslide debris. The
second category includes simplified analytical models,
which describe the physical behavior of debris
movement, based on lumped mass approaches in
which the debris mass is assumed as a single point.
The third includes numerical simulations of conservation
equations of mass, momentum and energy that
describe the dynamic motion of debris, and/or a
rheological model to describe the material behavior
of debris.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 71
4.2.1. Empirical models
The widely used empirical methods mainly
include mass-change method and the angle of reach.
The mass-change method is based on the phenomenon
that as the landslide debris moves downslope,
the initial volume/mass of the landslide is being
modified through loss or deposition of materials,
and that the landslide debris halts when the volume
of the actively moving debris becomes negligible
(Cannon and Savage, 1988). The average mass/
volume-change rate of landslide debris was established
by dividing the volume of mobilized material
from the landslide by the length of the debris trail.
The influence of slope gradient, vegetation types,
and the channel morphology on the average volumeloss
rate was established by stepwise multivariate
regression analysis. This method does not explicitly
account for the mechanics of the movement process.
Another empirical approach is the angle of reach,
defined as the angle of the line connecting the crest
of the landslide source to the distal margin of the
displaced mass. Scheidegger (1973) noted that the
angle of reach apparently decreases with increasing
magnitude of rock avalanches. This trend was first
documented for rock avalanches exceeding 0.1 to 1
million m3
. Corominas (1996) conducted a detailed
study on the influence of various factors that affect
the angle of reach using landslide records, and
showed a linear correlation between volume and
angle of reach for all types of failures. He found
that all kinds of mass movement show a continuous
decrease in angle of reach with increasing volume
starting from magnitudes as small as 10 m3
. He then
categorized the landslide records according to their
failure mechanisms into rockfalls, translational
slides, debris flows, and earth flows. Regression
equations for calculating the angle of reach of each
landslide type were developed, and it is clearly
indicated that the angle of reach decreases with an
increase in debris volume. Earth flows have the
highest mobility, and rockfalls have the lowest
mobility. However, a common problem with the
angle of reach method is that the scatter of the data
is too large to permit reliable use for any but the
most preliminary predictions of the travel distance.
Therefore, this method should be applied with some
judgment. For instance, if the material of a slope is
known to be dense and dilatant, larger angle of reach
will be applicable, than if the material is loose and
contractant.
Empirical methods are generally simple and relatively
easy to use, the information required by these
methods is usually general and readily available.
When a local historic landslide database is available,
the empirical relationships can be readily developed.
However, empirical methods can only provide a
preliminary estimate of the profile of the travel path.
4.2.2. Analytical methods
The analytical methods include different formulations
based on lumped mass approaches in which the
debris mass is assumed as a single point. The
simplest type of analytical methods is the sled model
(Sassa, 1988), which assumes that all energy loss
during debris movement is due to friction and describes
the landslide as a dimensionless body moving
down the profile of the path. Therefore, the movement
of a landslide is controlled by a single force
resultant, representing the gravity driving force as
well as all movement resistance. The ratio of the
vertical to horizontal displacement of the center of
gravity of the block equals the friction coefficient
used in the analysis. This method can provide an
effective means for the calculation of runout distance,
velocity and acceleration of debris movement.
However, this method is applicable only for smallscale
rockslides of limited displacements, which do
not disintegrate during motion. Sassa (1988) improved
the sled model by considering the effect of
pore fluid pressures at the sliding plane. He considered
the frictional resistance along the sliding plane
to be a function of the intrinsic internal friction angle
and the pore pressure coefficient, B. The B value can
be determined by either laboratory methods as suggested
by Sassa (1988) or back calculation from the
landslide cases. Thus, the apparent friction angle in
the improved sled model can be expressed as the
combined effects of the intrinsic internal friction
angle of debris material, and the motion-induced
pore pressure.
Hutchinson (1986) developed a model for the
prediction of runout distances of flowslides in loose,
cohesionless materials by assuming that the shape of a
debris flow is a uniformly spread out sheet. In the
model, the basal resistance of the debris mass is
assumed to be purely frictional, and the excessive
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8772
fluid pressure in the debris mass is assumed to be
dissipating according to the one-dimensional consolidation
theory. As debris moves downslope, the shear
resistance on the sliding plane increases due to a
dissipation of excessive pore pressure. The debris
mass halts when the resultant force along the sliding
plane becomes zero. Based on these assumptions, the
transient downslope movement of debris masses can
be calculated.
For large, deep-seated slow-moving landslides,
shear rate-dependent reduction/increase of residual
strength may be an important consideration. The
present knowledge of the effect of shear rate on
residual strength allows the formulation of a soil
model linking the strength on a pre-existing shear
zone with displacement and rate of displacement. This
model can be used in the study of stability and the
prediction of instability of slopes containing shear
zones near or at residual strength under static (i.e.
alteration of loading of the sliding mass and fluctuation
of the groundwater table) (e.g. Skempton et al., 1989;
Bracegirdle et al., 1991), and dynamic (i.e. earthquakes)
loading conditions (e.g. Lemos and Coelho,
1991).
The lumped mass cannot account for lateral confinement
and spreading of the flow and the resulting
changes in flow depth, and should therefore be
suitable only for comparing paths which are very
similar in terms of geometry and material properties.
To apply these methods, some specific parameters
are required, such as pore pressure parameters and
debris thickness, relation of residual strength with
shear rate. An assumption on the initial pore pressure
distribution is needed for the sliding-consolidation
method, while the apparent friction angle of the
debris along the sliding plane is needed for the sled
model.
4.2.3. Numerical methods
Numerical methods for modeling runout behavior
of landslide debris mainly include fluid mechanics
models and distinct element method. Continuum fluid
mechanics models utilize the conservation equations
of mass, momentum and energy that describe the
dynamic motion of debris, and a rheological model
to describe the material behavior of debris. The Bingham
rheological model is regarded as one of the most
well developed rheological models for describing the
flow properties of earth materials (Sousa and Voight,
1991; Chen and Lee, 2000). By solving a set of
governing equations with a selected rheological model
describing the flow properties of the debris, the velocity,
acceleration and runout distance of debris can be
predicted. Typical examples of those applications of
continuum models were given by Sousa and Voight
(1991), and Chen and Lee (2000). Generally, the
continuum fluid mechanics models are very sophisticated
and the rheological properties required are difficult
to determine. Hungr (1995) developed a modified
continuum model. The method is different from the
classical continuum fluid mechanics modeling methods
discussed above. In his model, debris mass is
discreted into column elements, and the rheological
properties of debris at the sliding plane are accounted
for. The longitudinal rigidity of the flowing mass is
considered in conjunction with the lateral earth pressure
coefficient. The friction resistance is assumed to
act only at the base of the sliding plane. Pore pressure
effects are incorporated with the pore pressure coefficient
—the ratio of pore pressure to the total normal
stress at the base of the column element. The friction
angle can be assumed to be a function of displacement
to simulate shear strength decays from peak to residual.
The pore pressure coefficient can be expressed as
a function of location or elapsed time to simulate
drainage and consolidation effects. This method is
easy to use, and the debris movements at different
time intervals can be simulated.
Continuum fluid mechanics models are very
sophisticated. Rheological models have to be selected,
and the required rheological parameters have to be
determined by either laboratory methods or by backanalysis
from the landslide cases with similar geological
conditions. However, they are able to provide
all the information required for landslide risk assess-
ment.
Another method commonly used for simulating
runout distance and velocity is the distinct element
method, which can be effectively used to model large
strain particle movement. This method is very sophisticated
and the analysis is very time-consuming.
However, it is an invaluable tool in understanding
the failure mechanics of landslides through back
analysis.
After determining the probability of landsliding
and the areal extent that would be potentially affected
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 73
by the landslide, landslide hazard can be delimited,
and elements at risk, which mean the population,
buildings, economic activities, public services utilities
and infrastructure in the area potentially affected by
landslides, can be readily defined.
5. Assessment of vulnerability
Vulnerability is a fundamental component in the
evaluation of landslide risk (Leone et al., 1996).
Vulnerability, in the present context, may be defined
as the level of potential damage, or degree of loss, of a
given element (expressed on a scale of 0 to 1)
subjected to a landslide of a given intensity (Fell,
1994; Leone et al., 1996; Wong et al., 1997). Vulnerability
assessment involves the understanding of the
interaction between a given landslide and the affected
elements. Generally, the vulnerability to landsliding
may depend on (a) runout distance; (b) the volume
and velocity of sliding; (c) the elements at risk
(buildings and other structures), their nature and their
proximity to the slide; and (d) the elements at risk
(persons), their proximity to the slide, the nature of the
building/road that they are in, and where they are in
the building, on the road, etc. (Finlay, 1996). The
vulnerability of lives and property to landsliding may
be different. For instance, a house may have similar
and high vulnerability to a slow-moving and a rapid
landslide, but persons living in the property may have
a low vulnerability to the slow-moving landslide but a
higher vulnerability to the rapid landslide (Fell, 1994;
Fell and Hartford, 1997).
The assessment of vulnerability is somewhat subjective
and largely based on historic records. For
example, the vulnerability of a house immediately at
the base of a steep slope down which a debris flow
may occur is clearly higher than for a house at the
limits of deposition area (because the velocity of flow
is much less) (Fell, 1994; Fell and Hartford, 1997).
Given a particular facility type and the probable depth
of debris at the facility location, the appropriate
vulnerability factor may be assessed systematically
by expert judgement. Another method for assessing
the vulnerability of person and/or property to landsliding
is based on the statistics of detailed historic
records. For instance, Hong Kong is virtually unique
in the world in terms of the detailed records kept on
landslides and their consequences. Finlay (1996)
presented an example where vulnerability is assigned
Table 1
Summary of Hong Kong vulnerability ranges and recommended values for death from landside debris in similar situations (from Finlay, 1996)
Case Vulnerability of person Comments
Range in data Recommended value
Person in open space
1. If struck by a rockfall 0.1–0.7 0.5a
May be injured
but unlikely
to cause death
2. If buried by debris 0.8–1.0 1.0 Death by asphyxia
3. If not buried 0.1–0.5 0.1 High chance of survival
Person in a vehicle
1. If the vehicle is buried/crushed 0.9–1.0 1.0 Death is almost certain
2. If the vehicle is damaged only 0–0.3 0.3 Death is highly likely
Person in a building
1. If the building collapses 0.9–1.0 1.0 Death is almost certain
2. If the building is inundated with debris
and the person buried
0.8–1.0 1.0 Death is highly likely
3. If the building is inundated with debris
and the person not buried
0–0.5 0.2 High chance of survival
4. If the debris strikes the building only 0–0.1 0.05 Virtually no dangera
a
Better considered in more detail, i.e. the proximity of person to the part of the building affected by sliding.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8774
by reference to historic data (Table 1). An alternative
option based on the statistics of historic records is the
vulnerability matrix method proposed by Leone et al.
(1996). This method is flexible and can cater for a
wide range of situations and can, to a certain degree,
reduce subjectivity, compared with the methods mentioned
above. With this method, the vulnerability of
elements at risk depends on the characteristics of the
landslide and the technical resistance of the building,
such as the type, nature, age, etc. For instance, Fig. 2
gives a correlation, in terms of vulnerability, between
exposed elements and the characteristics of landslides.
The applicability of this method, like other methods,
also requires statistical analysis of detailed records on
landslides and their consequences. When assessing
landslide risk, we must account for the indirect cost,
such as interruptions in economic activity, if we are to
calculate the risk in all of its aspects in real terms, in
addition to the direct damage caused by landsliding.
6. Landslide risk assessment
There are a variety of risks to be addressed in
landslide risk assessment, generally comprising distributed
landslide risk, site-specific landslide risk, and
global landslide risk.
6.1. Distributed landslide risk assessment
Distributed landslide risk assessment is aimed at
providing a risk map that depicts the level of risk in
terms of fatality or economic loss at different locations
of a given region quantitatively or qualitatively. The
spatial distribution of landslide risk may be obtained
by spatial subdivision of the area under study and
multiplication of spatial landslide probability, affected
zones, land-use or spatial distribution of population or
property, and vulnerability. This type of calculation
can easily be calculated within the GIS (Leone et al.,
1996).
A priority ranking system for selection of slopes
for detailed investigation and necessary upgrading,
which can systematically ranks old slopes in an order
of priority based on their probability of failure and the
consequence of failure, is a commonly used empirical
approach for risk assessment, based on an inventory
database of all potentially unstable slopes in an area.
For instance, the Geotechnical Engineering Office
(GEO) of the Hong Kong Government established
New Priority Classification Systems (NPCSs) for
slopes and retaining walls developed as part of the
GEO Slope Information System (SIS), under which a
total score can be calculated for soil cut slopes, rock
cut slopes, fill slopes and retaining slopes, reflecting
the relative risk of landslide involving the feature. The
total score is obtained from the multiplication of an
instability score and a consequence score. The instability
score is based on an assessment of a number of
key parameters that affect the likelihood of failure.
The consequence score reflects the likely consequence
of failure. The higher the total score, the higher the
priority for follow-up action on the feature generally
(Wong, 1998).
Fig. 2. An example of structural vulnerability matrix.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 75
6.2. Site-specific landslide risk assessment
Site-specific risk assessment serves to provide a
systematic assessment of the hazards and level of risk
in terms of fatality (or economic loss, as appropriate)
at a given site, or a potential landslide. This facilitates
the consideration of whether the risk levels are
acceptable and the evaluation of different risk mitigation
measures, usually on the basis of cost–benefit
analyses. In general, the event tree technique is well
suited to a site-specific quantitative risk assessment
given its greater degree of refinement. Once the
probability of landsliding, runout behavior of landslide
debris or elements at risk, and vulnerability
terms have been derived, risk values may be simply
obtained by Eqs. (1) or (2). Event-tree analysis can be
used to quantify the risk from the potential failure.
The procedure consists of the following steps (Wu et
al., 1996):
 Examine possible triggering factors, such as
earthquakes or/and rainstorms;
 Identify possible failure modes;
 Estimate probability of failure for each failure
mode;
 Evaluate runout behavior of landslide debris for
each failure mode;
 Assess the risk for each failure mode; and
 Sum the risk for all possible failure modes.
6.3. Global landslide risk assessment
A global risk assessment, on the other hand, is
aimed at defining the relative contribution to the total
risk (e.g. number of fatalities per year), which can
provide a reference for landslide risk management and
consideration of resources allocation and policy-making.
It can be calculated by summing site-specific risk
of all slopes in the region under study.
7. Landslide risk management
Once the risk from a landslide or areas susceptible
to or affected by landsliding are identified, measures
may be taken to mitigate landslide risk to the community
if necessary. The community faced with a
landslide has a variety of strategies to deal with it, and
these strategies may be grouped into planning control,
engineering solution, acceptance, and monitoring and
warning systems. Planning control may be seen as
reducing expected elements at risk; the engineering
solution strategy as reducing either the probability of
landsliding or the probability of spatial impact of a
landslide; the acceptance strategy as acceptable or
unavoidable; and the monitoring and warning system
strategy as reducing expected elements at risk by
evacuation in advance of failure. The risks assessed
as mentioned above can then be compared with the
acceptance criteria to decide upon the landslide mitigation
measures required.
7.1. Planning control
Planning control is one of the effective and economical
ways to reduce landslide losses. It can be
accomplished by (a) removing or converting existing
development, and/or (b) discouraging or regulating
new development in unstable areas (Kockelman,
1986). The latter option is the most economical and
effective means for local governments if feasible.
New developments can be prohibited, restricted or
regulated in landslide-prone areas. Landslide-prone
areas can be used as open space, parks, woodland
and recreation, or specific land-uses or operations that
might cause mass movement or that might be vulnerable
to failure, such as irrigation, construction of roads
and buildings, can be prohibited. In the US, excavation,
grading, and construction regulations have been
developed to ensure that construction in landslide
prone area is planned and constructed in a manner
that will not impair the stability of hillside slopes
(Schuster and Fleming, 1986; Schuster, 1996). These
commonly include regulating, minimizing, and prohibiting
excavation and fill activities in landslideprone
areas, requiring a permit prior to scraping,
excavating, filling or cutting any lands, and requiring
proper engineering design, construction, inspection,
and maintenance of cuts, fills, and drainage systems.
Landslide hazard maps at various scales are highly
valuable tools in assessing the landslide potential for
a proposed subdivision or parcel. They can also be
used as a basis for site-specific regulation, including
prohibition of development on active landslides,
requirement for detailed geotechnical analysis and
mitigation, and specific grading regulations.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8776
Recurrent damage from landslides can be avoided
by evacuating areas that continue to have slope failures.
This may be an effective management strategy
for reducing landslide risk for large-sized recurrent
renewed landslides. For large-scale potential landslides,
if there are obvious indications of instability,
evacuations must be considered.
7.2. Engineering solution
Engineering solution is the most direct and costly
strategy for reducing landslide risk. Two general
approaches are available for mitigating landslide risk.
One is correction of the underlying unstable slope to
control initiation of landslides, and the other is controlling
of the landslide movement.
7.2.1. Correction of the underlying unstable slope
The most commonly used remedial measures for
correction of the underlying unstable slope are modification
of the slope geometry by excavation or toe
fill, drainage of surface and ground water, use of
retaining structures, and internal slope reinforcement.
These remedial measures are excellent site-specific
management tools for landslides if correctly designed
and constructed. The major negative aspect in their
use is their high cost. For this reason, they are often
used where landslide costs are very high because of
high population densities and property values.
Slope geometry modification by removal of all or
part of the earth driving the landslide is the most
efficient way of increasing the factor of safety of a
slope, especially in deep-seated slides (Bromhead,
1997). The slope gradient can be flattened by removing
unstable material from the crest of the slope or/and
adding material to the toe to enhance the stability of a
slope. However, such an approach may not be easy to
adopt for long translational slides where there is no
obvious toe or crest, where the unstable area is so
complex that a change in topography to improve the
stability of one area may adversely affect the stability
of another.
Drainage is the most widely used method for slope
stabilization. Underground drainage systems and
pumping wells remove ground water and decrease
pore water pressure, thus increasing the shear strength
of soil. Surface water is drained from the unstable
areas by surface ditches so as to reduce surface water
infiltration into the potential slide mass. In the design
of underground drainage systems to stabilize slopes,
the capacity of each drain or outlet pipe should be
accounted for. The underground drainage system
should have the capacity to decrease pore water
pressure at the failure surface in order to stabilize
the landslide as much as possible.
Slope stability can also be increased by placing
retaining structures to increase the resistance to
movement. These include gravity retaining walls,
crib-block walls, gabion walls, passive piles, piers
and caissons, cast-in situ reinforced concrete walls,
reinforced earth-retaining structures with strip/sheetpolymer/metallic
reinforcement elements, etc. Generally,
the following minimum information is required
to determine the type and size of a restraining
structure: (a) the boundaries and depth of the unstable
area, its moisture content and its relative stability;
(b) the type of landslide that is likely to develop or
has occurred; and (c) the foundation conditions, since
restraining structures require a satisfactory anchorage.
In addition, the site conditions for construction
may also affect the choice of retaining structures.
Internal slope reinforcement is aimed at modifying
the fundamental properties of in-situ slope soils to
improve the internal resistance against sliding by
block bolts, micropiles, soil nailing, anchor, grouting
with cement or chemicals, stone or lime/cement
columns.
7.2.2. Controlling of the landslide movement
An alternative landslide risk-mitigating strategy of
engineering solutions is to control the movement of
landslide debris so as to reduce the spatial impact of
landslides on downslope property. Traditional mitigation
measures address landslide hazards through
the installation of mechanical barriers where necessary
to protect structures. The most frequently used
designs include diversionary structures or levees to
direct landslide debris into predetermined depositional
areas, debris defenses intended to absorb kinetic
energy, and retaining walls designed to withstand
and/or deflect impacts (Montgomery et al., 1991).
However, the volume of a potential landslide likely
to impact a specific site should be estimated qualitatively
or quantitatively. Knowledge of material
properties of landslide debris and topography below
the landslide site could be used to estimate the mass
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 77
and velocity of landslide debris likely to impact a
site as an input to the predictive models of runout
distance outlined previously. Debris defenses on
hillslopes below the landslide sources typically consist
of a series of post set in concrete and connected
by either wooden cross members or a chain-link
fence. Such structures might decelerate small failures,
but they are relatively ineffective on very steep
slopes or for containing large failures, and thus
should not be employed to protect structures (Montgomery
et al., 1991). Retaining walls can be
installed downslope of landslides to stop moving
landslide debris. However, care should be taken to
consider their effectiveness and economics because
they are vulnerable to overtopping by either high
velocity landslide debris or accumulated deposits
from repeated events. As well, they are relatively
expensive to build, requiring specific design considerations
for both the volume and velocity of landslide
events.
7.3. Acceptance
A community may need to accept the risk from a
given landslide under the condition that the risks are
perfectly understood. The selection of this option is
based on considerations that the risks from landslides
are offset against the benefits which the
community obtains in the particular locations, or
that risks from landslides are tolerable (Bromhead,
1997).
Before the acceptance strategy is adopted, acceptable
risk criteria should be established. Fell (1994)
and Fell and Hartford (1997) provide a summary of
the issues that affect establishing levels of tolerable
risk. IUGS Working Group on Landslides, Committee
on Risk Assessment (1997) defines the tolerable risk
as a risk that society is willing to live with so as to
secure certain benefits in the confidence that it is
being properly controlled, kept under review and
further reduced as and when possible, and provides
the following general principles that can be applied
when considering tolerable risk criteria:
 The incremental risk from a hazard to an
individual should not be significant compared
to other risks to which a person is exposed in
everyday life;
 The incremental risk from a hazard should,
wherever reasonably practicable, be reduced,
i.e. the As Low As Reasonably Practicable
(ALARP) principle should apply;
 If the possible loss of life from a landslide
incident is high, the risk that the incident might
actually occur should be low. This accounts for
the particular intolerance of a society to
incidents that cause many simultaneous casualties,
and is embodied in societal tolerance risk
criteria;
 Persons in the society will tolerate higher risks
than they regard as acceptable, when they are
unable to control or reduce the risk because of
financial or other limitations;
 Higher risks are likely to be tolerated for
existing slopes than for planned projects, and
for workers in industries with hazardous slopes,
e.g. mines, than for society as a whole;
 Tolerable risks are higher for naturally occurring
landslides than those from engineered slopes;
 Once a natural slope has been placed under
monitoring, or risk mitigation measures have
been executed, the tolerable risks approach
those of engineered slopes;
 Tolerable risks may vary from country to
country, and within countries, depending on
historic exposure to landslide hazard, and the
system of ownership and control of slopes and
natural landslide hazards.
In theory it is desirable to establish such criteria
and develop the F–N curve for landslide hazard. The
F–N curve is a plot showing the number of fatalities
(N) plotted against the cumulative frequency ( F ) of N
or more fatalities on a log–log scale (Wong et al.,
1997; Ho et al., 2000). Tolerable risk criteria also
usually require that maximum involuntary risk of
death to a single individual in a specified location
should not exceed a specified threshold (Morgenstern,
1997). Establishing such criteria for landslide hazards
should make use of existing data to construct F–N
curves. The concept in each case would be to gather
data on the frequency of landsliding causing fatalities
or juries, and analyzing this for the population of
slope features, and the reaction of the public to the
fatalities or juries, and the measures taken by the
owners of the slopes, or government authorities, to
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8778
mitigate the risk. However, Morgenstern (1997) stated
that few organizations have a rich enough database to
progress very far in this direction, with an exception
of Hong Kong. Under this condition, two approaches
may possibly be used. First, event tree models of
landslide failure and consequence based on conditional
probabilities can be constructed and used to
determine theoretical F–N relations (e.g. Wong et al.,
1997). However, careful calibration wherever possible
is essential if this approach is to lead to credible
results. Secondly, social and political judgements
can be used on a case-by-case basis to help determine
acceptable risk (e.g. Bunce et al., 1995).
Defining tolerable risk is a complex problem that
varies for individuals and societies. Using an agreed
figure for acceptable risk as a guide, risk contour map
can be produced and planning-decisions based on
them. In Hong Kong, an interim risk guideline (Geotechnical
Engineering Office, 1998) was issued for
landslides and boulder falls from natural terrain. The
maximum allowable individual risk was set at 10 À 5
for new developments, and 10 À 4
for existing developments.
For societal risk, the preferred criteria have
no acceptable line on the F–N curve, and the principle
of ALARP should be applied to all risks below the
unacceptable line. A region of intense scrutiny is set
between 1000 and 5000 fatalities. The recommendation
is used as a guideline and not meant to be
mandatory. An alternate option, by adding a broadly
acceptable limit of 10 À 5
for a fatality and 10 À 8
for
1000 fatalities, was proposed (Fig. 3). It may be
rational to accept the risk if the calculated risk from
a specified potential landslide is below the acceptable
risk. For societal risk, it is important to recognize that
there is a degree of uncertainty in the analysis of risk,
and that the individual and societal risk criteria are
only a mathematical expression of the tolerance of
society to risk. They are not precise, and should be
used only as a general guide.
7.4. Monitoring and warning systems
Potentially unstable slopes can be monitored so
that the potentially affected residents can be warned
and, if necessary, evacuated. For specific landslides or
potential slopes of such a large magnitude that stabilization
works or engineering solutions were not only
impracticable but would also not be cost-effective in
relation to the property in risk, monitoring and landslide
warning would be an alternative option to reduce
landslide risk. The response of a particular slope when
subjected to an earthquake is generally examined
using the critical acceleration under which the slope
would be brought to a state of limit equilibrium
according to a pseudo-static analysis. When the acceleration
imposed on the slope exceeds its critical
acceleration, displacement or failure will result. A
warning can be issued if an earthquake that can
produce peak ground acceleration exceeding its critical
acceleration can be predicted in advance. If a
Fig. 3. Interim societal risk in Hong Kong: (a) preferred option; (b) alternate option.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 79
slope is most possibly to be brought to failure by
rainfall, a monitoring system can be installed, and
landslide warning based on real-time monitoring data
may be issued.
The primary objective of a monitoring program for
a particular slope or landslide is to assess the existing
conditions, in particular to determine whether or not
the landslide is active, and if it is, where it is moving
and the rate at which it is moving, so as to warn of
impending emergencies. The monitoring instruments
should be located and installed with reference to the
geological conditions such that the overall picture of
slope movement can be defined. Monitoring typically
covers: (a) magnitude, rate, location and direction of
deformations, including surface deformations by using
crack gauges/surface extensometers, GPS or multipoint
liquid level gauges, and subsurface deformations
by using inclinometers, borehole/slope extensometers,
or multiple deflectometers; (b) pore pressures and
piezometric levels by using piezometers and tensiometers,
hydro-meteorological parameters by using rain
gauges; and (c) seismic accelerations if necessary. In
addition, the general fact that slope movements in rock
and soil masses are accompanied by the generation of
acoustic emission or noise makes it possible to detect
acoustic emission as an indication of landslide realtiming
warning (Kousteni et al., 1999). In selecting the
method of monitoring and the type of instrumentation,
the required accuracy and durability in relation to the
available time and the expected rate of deformation
have to be considered. For instance, a borehole inclinometer
will precisely detect minor deformations but its
lifetime will be very limited when it traverses an active
failure surface. Alternatively a downhole slope extensometer
will not give quite the same information and
not with the same sensitivity and accuracy but it may
endure large displacements. Recently, satellite-borne
synthetic aperture radar (SAR) interferometry (InSAR)
has been a powerful tool for detecting and monitoring
mass movements of unstable slopes (e.g. Rott et al.,
1999; Kimura and Yamaguchi, 2000). The main contribution
of InSAR to slope monitoring is that InSAR
produces three-dimensional maps of the earth’s surface,
based on the evaluation of phase difference of
two SAR images (interferogram) of the same area
using a couple of satellites on different, albeit similar,
orbits. A necessary condition for generating an interferogram
is that the electromagnetic characteristics of
the scene must stay constant between the two acquisitions.
Differential SAR interferometry enables
researchers to evaluate any variations over time of
the altimetric profile of the area under observation and
then to monitor slow movements with typical displacements
on the order of centimeters (Rott et al., 1999).
However, for slowly sliding surfaces interferograms
over long time spans are needed to obtain detectable
phase shifts. This prevents the application over densely
vegetated surfaces due to decorrelation of the SAR
images. Other constraints are imposed by the SAR
illumination geometry and the availability of repeat
pass interferometric data. The ability of InSAR to
provide continuous maps of surface deformation is a
considerable advantage over conventional techniques.
Other advantages are the global access, the synoptic
observation of large areas, and the regular repeat
coverage.
Based on the monitoring data, the performance of a
slope can be assessed. If the assessment indicates that
the landslide is active, or potentially unstable, three
strategies as mentioned above are applicable. The first
one is to do nothing and accept the consequences of
failure. Secondly, the slope can be stabilized and a
monitoring program installed can be used to verify the
effectiveness of stabilization works. Thirdly, a monitoring
program can be used to warn of instability so
that evacuation can be carried out prior to failure
occurring.
Another important area of landslide research
involves the development of real-time warning systems
for issuing landslide warnings in large areas
during major storms. Such a system is generally based
on (a) empirical and theoretical relations between
rainfall characteristics and landslide initiation; (b) geological
determination of areas prone to landslides; (c)
real-time monitoring of a regional network of telemetering
rain gauges; and (d) precipitation forecasts
made by the weather service authorities (Keefer et al.,
1987). The U.S. Geological Survey in cooperation
with the U.S. National Weather Service developed a
real-time landslide warning system for the San Francisco
Bay Region, California, and this system was
used to issue the first regional public warnings over
public radio and television stations during the storms
of 12–21 February 1986, which produced 800 mm of
rainfall in the San Francisco Bay Region. According to
eyewitness accounts of landslide occurrence, the warnF.C.
Dai et al. / Engineering Geology 64 (2002) 65–8780
ings accurately predicted the times of major landslide
events (Keefer et al., 1987). In Hong Kong, the Geotechnical
Engineering Office (GEO) also relies on a
rainfall-monitoring system and an empirical rainfall/
landslide relation for issuing regional landslide warnings
to the public (Brand et al., 1984).
The relationship between landslide occurrence and
rainfall characteristics forms the basis for the development
of regional real-time landslide warning systems.
Many attempts have been made to establish relationships
between rainfall and landslides. These relationships
are generally based on the assumption that there
exists a direct relationship between the occurrence of
landslides and the quantity of rainfall, in terms of
rainfall intensity and duration of rainstorm events, or
short-term rainfall (e.g. 24-h rainfall) and antecedent
rainfall. For instance, Caine (1980) collected a worldwide
set of rainfall data recorded near reported sites of
landslide occurrences and derived a rainfall/landslide
threshold by fitting a lower bound to a log–log plot of
peak intensities versus duration. Similar rainfall intensity
and duration thresholds were developed, based on
historical rainfall events, for the San Francisco Bay
Region (Cannon and Ellen, 1985), for the highly
susceptible La Honda region in the central Santa Cruz
Mountains, California (Wilson and Wieczorek, 1995),
and in the central mountains of Puerto Rico (Larsen
and Simon, 1993), respectively. In Hong Kong, Brand
et al. (1984) noted that a rainfall intensity of about 70
mm/h appeared to be the threshold value above which
landslides that resulted in casualties occurred, and that
landslides are unlikely for a rainfall of less than 100
mm in 24 h, but are almost certain when 175 mm is
exceeded.
In establishing a rainfall/landsliding relation for a
particular region, there are debates over the role of the
antecedent rainfall in landslide occurrence. Previous
studies in certain parts of the world (e.g. Lumb, 1975;
Wieczorek, 1987; Wilson and Wieczorek, 1995) on
the relationship between rainfall and landslides have
concluded that both the antecedent rainfall and a
critical intensity of rainfall are equally important
factors in triggering landslides. The significant period
of antecedent rainfall, however, may vary from hours
to weeks depending on local site conditions, particularly
the soil permeability of slopes and the depth of
failure. In the case of highly permeable soils in most
tropical areas where the failure surface of landsliding
is generally shallow, the period of necessary antecedent
rainfall is considered to be very short, and
antecedent rainfall is therefore not an important factor
(Brand et al., 1984).
Another issue is the rainfall parameter used for
establishing rainfall/landsliding relations. Wieczorek
(1987) examined landslides initiated in storms of
different intensity and duration in a 10-km2
area near
La Honda, California, and concluded that deep slides
in soils usually occurred as a result of long duration,
moderate intensity storms, whereas very shallow
slides of soil over bedrock occurred due to short
duration, high intensity storms. A similar observation
was made in Puerto Rico by Larsen and Simon
(1993). This indicates that the relationship between
rainfall and landslides may be failure-volume dependent.
Dai and Lee (2001) studied the relationship
between rainfall and the number of landslides in Hong
Kong, and found that with an increase in failure
volume of landslides, the most important rainfall
variable may vary from rainfall of short duration
(12-h rolling rainfall) to that of relatively long duration
(24-h rolling rainfall).
A lengthy period of data on landslides and rainfall
is necessary for establishing reliable rainfall/landslide
relationships. For instance, a study carried out by Dai
and Lee (2001) indicates that the period of historic
data has a great influence on the analytic results, and
that the rainfall/landsliding relation should be
updated, as new data become available.
Regional warning systems could alert the general
public to the potential landslide activity, although the
amount and intensity of rainfall necessary to initiate
landslides undoubtedly varies with site specific geological,
hydrological, and soil or rock properties. Such
systems can only provide information on when landsliding
would occur, and are thus most useful if used
in conjunction with regional landslide hazard mapping
to both delineate potentially hazardous areas and
determine the timing of landslide initiation.
7.5. Decision-making
The choice between available options is dependent
on the preference of the decision-maker considering
all possible outcomes of each of the options. After the
probabilities and consequences have been estimated,
methods of decision analysis may be used to arrive at
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 81
management decisions by identifying the alternatives
of actions, possible outcomes, and respective consequences
or cost for each scenario. The alternative with
the least expected cost is usually chosen if the
expected value is the criterion for decision. Cost–
benefit analysis is the most widely used method in the
process of decision-making. It involves the identification
and quantification of all desirable and undesirable
consequences of a particular mitigation measure.
When measuring the cost of risk, a monetary value is
generally used. By identifying the various available
options together with relevant information for assigning
design parameters, the cost and benefit of each
mitigation measure can be assessed.
8. Discussion
Landslide risk assessment and management has led
to the development of a new paradigm which demands
a pluralistic approach to risk assessment and risk
management and for value-focused decision making. It
embeds the use of science in risk assessment and risk
management and landslide studies within a sociopolitical
framework, and subsumes within the decision
making process the nature of landslides and the nature
of human activities. This is not to diminish the role
of scientists and of engineers in decision-making in
landslide hazard mitigation. On the contrary, the new
paradigm relies crucially on a better understanding of
landslide processes and on greater sophistication, transparency
and rigor in the application of science, but
within a collaborative and consensual decision-making
framework.
However, in the present knowledge, there are few
reliable techniques available for assessing landslide
hazard that is defined as the probability of occurrence
of a given magnitude of failure. Those available often
require detailed geotechnical information on the existing
conditions. Because of the high cost involved they
are generally only achievable at the site investigation
level in cases where high risk is anticipated. Probability
of landsliding and runout behavior of landslide
debris are more commonly assessed by historically
based, largely stochastic analysis of precedents. The
accuracy with which landslide probability can be
determined thus depends largely on the length, quality
and nature of the information record. Probability of
occurrence can be determined either directly from
landslide record or indirectly by establishing the landslide-triggering
threshold of the initiating agent and
analyzing agent behavior. With both approaches, the
minimum information requirements are a record of
events including descriptions to characterize location,
date, associated triggering conditions, causes, nature
and length of warning, as well as landslide magnitude
characteristics such as volume, material type and nature,
and extent of runout. The main drawback of the
precedent approach is the uncertainty associated with
applying the findings to areas beyond where the precedence
was established. For this reason or because there
is insufficient record of past activity, hazard assessments
are sometimes based on theoretically determined
causative factors rather than on precedence. The outcome
of this approach is the ranking of terrain susceptibility
to landsliding. The validation of landslide
susceptibility mapping and its usefulness depends on
the maintenance of appropriate records indicating the
magnitude and frequency of on-going landslide activity
and its relationship with terrain and triggering condi-
tions.
Although it is still only possible to predict slope
failure in most general terms and virtually impossible
to forecast the location, magnitude and timing of
specific future events, regional scale landslide risk
studies could result in the identification of tracts of
land with different levels of hazard and risk. Such a
hazard and risk zoning provides an ideal framework
for land-use planning, development control, the application
of building codes/ordinances, guidelines for
engineering practice. At large and site-specific scales,
landslide risk studies can provide the information
necessary for decision-making. The process of landslide
risk assessment and management depends on the
completeness and quality of basic information on
historic landslide data, and other physical and social
data. In developing an inventory of all landslides in
the area under study, one should determine the ages of
different landslide deposits, with the aim of estimating
recurrence intervals and determining risk where possible,
and incorporate landslide trigger data, such as
rainfalls and earthquakes, to establish time-frequency
probabilities for singe major storms and for highrainfall
seasons. It is very clear that limited historic
data often constrains the use of vulnerability assessment
and risk analysis in land-use planning.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8782
To move toward the goal that landslide risk can be
rapidly assessed and effectively managed, a comprehensive
landslide information system is required. In
this regard, Geographical Information Systems (GIS),
linked with remote sensing technology and telecommunications/warning
systems, have emerged as one of
the most promising tools to support the landslide risk
assessment and management process. Establishment
of a landslide database is the basis for landslide risk
assessment and management. The landslide records
generally include geotechnical, lithological, geomorphologic
information, the date and extent of failure,
and consequence from individual landslide sites. Data
on existing engineered slopes, which may have influence
on property in the event of failure, are also
required. These data include slope geometry, evidence
of past instability, sign of distress, slope-forming
material, age and status of slope protective measures,
site investigation and laboratory test data, which may
be needed for assessing the priority of follow-up
measures. Physical and social data are needed for all
subsequent probability, vulnerability and risk assessment
and management. These data should include
topographic, geological, seismological and geomorphologic
data at various scales. A further requirement
is social, economic, cultural, political and environmental
factors so as to determine elements at risk for
the purpose of vulnerability assessment. Real-time
monitoring data on rainfall, earthquake, pore pressure,
and displacement of landslides, which may be a
trigger for landslide occurrence, or indications of the
movement of specific landslides, should be communicated
into the database in real-time for the purpose
of landslide warning. This information can be
acquired at sites by local monitoring units (LMUs),
which are directly connected to the central network
monitoring, or by remote monitoring units (RMUs),
which transmit data to the central network monitor by
telemetry (usually radio) links. The individual RMU
or LMU systems contain modules for the switching,
signal conditioning, power excitation, control, and
communications functions required to acquire data
from each transducer. RMU and LMU systems can
be designed to convert raw data readings to the desired
engineering unit, to collect data readings according
to a pre-programmed sequence, or to be polled
from the central network module. The development
of the computer-based automated data analysis systems
(ADAS) is generally capable of dealing with a
variety of transducer types and instrument monitoring
needs.
Once the spatial database has been developed, the
manipulation and analysis of the data allow it to be
combined in various ways to evaluate what will
happen in certain situations. The overlay operation of
the GIS coupled with both heuristic and statistical
approaches allows us to combine factor maps of site
characteristics in a variety of ways to produce susceptibility
maps (Gupta and Joshi, 1989; Carrara et al.,
1991; van Westen et al., 1997; Chung and Fabbri,
1999). At large and site-specific scales, process
models are in use that simulate the spatial distribution
of the factor of safety using slope stability models
(e.g. Wu and Sidle, 1995; van Westen et al., 1997).
Most present applications of GIS to landslide studies
are based on map overlaying, which only allows the
spatial comparison of different maps at common pixel
locations. A relatively new development in the use of
GIS for slope instability assessment is the application
of the so-called ‘‘neighborhood analysis’’ (e.g. Wu
and Sidle, 1995). Neighborhood operations permit
evaluation of the neighboring pixels surrounding a
central pixel, which can be used to simulate trajectories
of paths for slope processes and hydrological
response to rainfall as one of the input parameters in
infinite slope modeling (Wu and Sidle, 1995). Temporal
database information is correlated with historical
triggering factors to calculate temporal probabilities
for landslide forecasting (Dikau et al., 1996).
However, the above-mentioned specific capabilities
desired to perform risk analysis, risk assessment and
management in a spatial environment are not all
available in a standard off-the-shelf GIS. It is desirable
to design a system that would integrate the capabilities
of selected components and programs. In designing
and developing such a system for landslide risk assessment
and management, an architecture consisting of
interconnected subsystems is selected. The subsystems
encompass aspects of the overall system that share
some common property or have similar functionality.
Examples of subsystems in the development described
herein are third-party software packages as well as
original programs and macro languages. Several prototypical
architectural frameworks are common in
existing systems, and the one adopted for this system
may be an interactive interface. This type of interface
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 83
is dominated by the interactions between external
software packages and the system, such that the
control of the analysis remains with the user who, in
this case, is expected to have some knowledge of both
geotechnical engineering and spatial analysis technology.
Present knowledge on the regional and local
landslide risk assessment indicates that statistical analysis
should be available to regional and local landslide
susceptibility assessment. For this purpose, commercial
statistical software packages can be integrated into
the system. For specific slopes, commonly used geotechnical
software on slope stability analysis, groundwater
modeling, and runout modeling of landslide
may be integrated with the system. Based on the geotechnical
data available for a specific slope, detailed
geotechnical analysis can be performed with the data.
Other software for assessing slope performance based
on monitoring data, priority classification systems for
slopes for the purpose of follow-up measures, should
be developed since apparently no commercial software
packages are presently available for such pur-
poses.
As shown in Fig. 4, all of the aforementioned
techniques can generally be integrated for the purpose
of landslide risk assessment and management.
9. Concluding remarks
Significant progress has been made in the field of
landslide risk assessment. The availability and quality
of historic landslide database cannot be overemphasized
since they constitute the basis for all components
of landslide risk assessment. Modern technologies,
such as GIS and remote communications, should have
a wider application in landslide risk assessment and
management.
Acknowledgements
Funding for this research was provided by the
Research Grants Council of Hong Kong and the
Hong Kong Jockey Club Research and Information
Fig. 4. A GIS-based conceptual integrated system for landslide risk assessment and management.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8784
Centre for Landslip Prevention and Land Development,
at the University of Hong Kong. The authors
wish to express their sincere appreciation for the
generous support received from these two organiza-
tions.
References
Bracegirdle, A., Vaughan, P., High, D., 1991. Displacement prediction
using rate effects on residual shear strength. In: Bell, D.H.
(Ed.), Landslides, vol. 1. Balkema, Rotterdam, pp. 343–347.
Brand, E.W., Premchitt, J., Phillipson, H.B., 1984. Relationship
between rainfall and landslides. Proc. 4th International Symposium
on Landslides, Toronto, vol. 1. BiTech Publishers, Vancouver,
Canada, pp. 377–384.
Bromhead, E.N., 1997. The treatment of landslides. Proceedings of
Institution of Civil Engineers, Geotechnical Engineering 125,
85–96.
Bunce, C.M., Cruden, D.M., Morgenstern, N.R., 1995. Hazard assessment
for rock fall on a highway. Proceedings of the 48th
Canadian Geotechnical Conference, Vancouver, 449–508.
Caine, N., 1980. The rainfall intensity–duration control of shallow
landslides and debris flows. Geografiska Annaler 62A, 23–27.
Cannon, S.H., Ellen, S.D., 1985. Rainfall conditions for abundant
debris avalanches, San Francisco Bay region, California. California
Geology 38 (12), 267–272.
Cannon, S.H., Savage, W.Z., 1988. A mass change model for debris
flow. The Journal of Geology 96, 221–227.
Carrara, A., Cardinali, M., Detti, R., Guzzetti, F., Pasqui, V., Reichenbach,
P., 1991. GIS techniques and statistical models in evaluating
landslide hazard. Earth Surface Processes and Landforms
16, 427–445.
Chen, H., Lee, C.F., 2000. Numerical simulation of debris flows.
Canadian Geotechnical Journal 37, 146–160.
Chung, C.F., Fabbri, A.G., 1999. Probabilistic prediction models for
landslide hazard mapping. Photogrammetric Engineering and
Remote Sensing 65 (12), 1389–1399.
Corominas, J., 1996. The angle of reach as a mobility index for
small and large landslides. Canadian Geotechnical Journal 33,
260–271.
Cruden, D.M., 1991. A simple definition of a landslide. Bulletin of
the International Association of Engineering Geology 43, 27–
29.
Dai, F.C., Lee, C.F., 2001. Frequency–volume relation and prediction
of rainfall-induced landslides. Engineering Geology 59
(3/4), 253–266.
Dai, F.C., Lee, C.F., Wang, S.J., Feng, Y.Y., 1999. Stress–strain
behavior of a loosely compacted volcanic-derived soil and its
significance to fill slope failures. Engineering Geology 53,
359–370.
Dai, F.C., Chen, S.Y., Lee, C.F., 2000. Analysis of the mechanism
of landslide initiation based on stress–strain behavior of soils.
Chinese Journal of Geotechnical Engineering 22 (1), 127–130
(in Chinese).
Dikau, R., Cavallin, A., Jager, S., 1996. Databases and GIS for landslide
research in Europe. Geomorphology 15 (3/4), 227–239.
Fell, R., 1994. Landslide risk assessment and acceptable risk. Canadian
Geotechnical Journal 31, 261–272.
Fell, R., Hartford, D., 1997. Landslide risk management. In: Cruden,
D., Fell, R. (Eds.), Landslide Risk Assessment. Balkema,
Rotterdam, pp. 51–109.
Fell, R., Chapman, T.G., Maguire, P.K., 1991. A model for prediction
of piezometric levels in landslides. In: Chandler, R.J. (Ed.), Slope
Stability Engineering. Thomas Telford, London, pp. 73–82.
Finlay, P.J., 1996. The risk assessment of slopes. School of Civil
Engineering, University of New South Wales, Australia, PhD
thesis.
Fleming, R.W., Ellen, S.D., Algus, M.A., 1989. Transformation of
dilative and contractive landslide debris into debris flows — an
example from Marin County, California. Engineering Geology
27, 201–223.
Geotechnical Engineering Office, 1998. Landslides and Boulder
Falls from Natural Terrain: Interim Risk Guidelines. GEO Report
No. 75. Geotechnical Engineering Office, Hong Kong,
183 pp.
Gupta, R.P., Joshi, B.C., 1989. Landslide hazard zoning using the
GIS approach — a case study from the Ramganga catchment,
Himalayas. Engineering Geology 28, 119–131.
Ho, K., Leroi, E., Roberts, B., 2000. Keynote lecture: quantitative
risk assessment — application, myths and future direction. Proceedings
of the International Conference on Geotechnical and
Geological Engineering, GEOENG 2000, Melbourne, Australia,
vol. 1. Technomic, Lancaster, pp. 236–312.
Hungr, O., 1995. A model for the runout analysis of rapid flow
slides, debris flows, and avalanches. Canadian Geotechnical
Journal 32, 610–623.
Hungr, O., Yau, H.W., Tse, C.M., Cheng, L.F., Hardingham, A.D.,
1999. Natural slope hazard and risk assessment framework. In:
Clarke, B. (Ed.), Urban Ground Engineering. Thomas Telford,
London, pp. 332–353.
Hutchinson, J.N., 1986. A sliding-consolidation model for flow
slides. Canadian Geotechnical Journal 23, 115–126.
Hutchinson, J.N., 1995. Keynote paper: landslide hazard assessment.
In: Bell, D.H. (Ed.), Landslides. Balkema, Rotterdam,
pp. 1805–1841.
IUGS Working Group on Landslides, Committee on Risk Assessment,
1997. Quantitative risk assessment for slopes and landslides
— the state of the art. In: Cruden, D., Fell, R. (Eds.),
Landslide Risk Assessment. Balkema, Rotterdam, pp. 3–12.
Keefer, D.K., Wilson, R.C., Mark, R.K., Brabb, E.E., Brown III,
W.M., Ellen, S.D., Harp, E.L., Wieczorek, G.F., Alger, C.S.,
Zatkin, R.S. 1987. Real-time landslide warning during heavy
rainfall. Science 238, 921–925.
Kimura, H., Yamaguchi, Y., 2000. Detection of landslide areas using
satellite radar interferometry. Photogrammetric Engineering
and Remote Sensing 66 (3), 337–344.
Kockelman, W.J., 1986. Some techniques for reducing landslide
hazards. Bulletin of the Association of Engineering Geologists
23 (1), 29–52.
Kousteni, A., Hill, R., Dixon, N., Kavanagh, J., 1999. Acoustic
emission technique for monitoring soil and rock slope instabilF.C.
Dai et al. / Engineering Geology 64 (2002) 65–87 85
ity. In: Yagi, N., Yamagami, T., Jiang, J.C. (Eds.), Slope Stability
Engineering. Balkema, Rotterdam, pp. 150–156.
Larsen, M.C., Simon, A., 1993. A rainfall intensity–duration
threshold for landslides in a humid–tropical environment, Puerto
Rico. Geografiska Annaler 75A (1–2), 13–23.
Lemos, L.J.L.J., Coelho, P.A.L.F., 1991. Displacements of slopes
under earthquake-loading. Proceedings of 2nd International
Conference on Recent Advances in Geotechnical Earthquake
Engineering and Soil Dynamics, Rolla, MO. Balkema, Rotterdam,
pp. 1–6.
Leone, F., Aste, J.P., Leroi, E., 1996. Vulnerability assessment of
elements exposed to mass-moving: working toward a better risk
perception. In: Senneset, K. (Ed.), Landslides. Balkema, Rotterdam,
pp. 263–269.
Leroueil, S., Locat, J., 1998. Slope movements—geotechnical characterization,
risk assessment and mitigation. In: Maric, B., Lisac,
L., Szavits-Nossan, A. (Eds.), Geotechnical Hazards. Balkema,
Rotterdam, pp. 95–106.
Leroueil, S., Locat, J., Vaunat, J., Picarelli, L., Lee, H., Faure, R.,
1996. Geotechnical characterization of slope movements. In:
Senneset, K. (Ed.), Landslides. Balkema, Rotterdam, pp. 53–74.
Li, T., Wang, S., 1992. Landslide Hazards and their Mitigation in
China. Science Press, Beijing, 84 pp.
Lumb, P., 1975. Slope failures in Hong Kong. Quarterly Journal of
Engineering Geology 8, 31–65.
Montgomery, D.R., Wright, R.H., Booth, T., 1991. Debris flow
hazard mitigation for colluvium-filled swales. Bulletin of the
Association of Engineering Geologists 28 (3), 303–323.
Morgan, G.C., Rawlings, G.E., Sobkowicz, J.C., 1992. Evaluating
total risk to communities from large debris flows. Proceedings
of 1st Canadian Symposium on Geotechnique and Natural Hazards.
BiTech Publishers, Vancouver, BC, Canada, pp. 225–236.
Morgenstern, N.R., 1997. Toward landslide risk assessment in practice.
In: Cruden, D., Fell, R. (Eds.), Landslide Risk Assessment.
Balkema, Rotterdam, pp. 15–23.
Mostyn, G.R., Fell, R., 1997. Quantitative and semiquantitative
estimation of the probability of landsliding. In: Cruden, D., Fell,
R. (Eds.), Landslide Risk Assessment. Balkema, Rotterdam, pp.
297–315.
Nicoletti, P.G., Sorriso-Valvo, M., 1991. Geomorphic controls of the
shape and mobility of rock avalanches. Geological Society of
America Bulletin 103, 1365–1373.
Rott, H., Scheuchl, B., Siegel, A., Grasermann, B., 1999. Monitoring
very slow slope movements by means of SAR interferometry:
a case study from a mass waste above a reservoir in the
Otztal Alps, Austria. Geophysical Research Letters 26 (11),
1629–1632.
Sassa, K., 1988. Geotechnical model for the motion of landslides. In:
Bonnard, C. (Ed.), Proceedings of 5th International Symposium
on Landslides, Lausanne. Balkema, Rotterdam, pp. 37–55.
Scheidegger, A., 1973. On the prediction of the reach and velocity
of catastrophic landslides. Rock Mechanics 5, 231–236.
Schuster, R.L., 1996. Socioeconomic significance of landslides. In:
Turner, A.K., Schuster, R.L. (Eds.), Landslides: Investigation
and Mitigation, Special Report 247, Transportation Research
Board, National Research Council. National Academy Press,
Washington, DC, pp. 12–35.
Schuster, R.L., Fleming, R.W., 1986. Economic losses and fatalities
due to landslides. Bulletin of the Association of Engineering
Geologists 23 (1), 11–28.
Schuster, R.L., Leighton, R.W., 1988. Socioeconomic significance
of landslides and mudflows. In: Kozlovskii, E.A. (Ed.), Landslides
and Mudflows. UNESCO/UNEP, Moscow, pp. 131–141.
Skempton, A.W., Leadbeater, A.D., Chandler, R.J., 1989. The Mam
Tor landslide, north Derbyshire. Philosophical Transactions of
the Royal Society of London A 329, 503–547.
Slosson, J.E., Krohn, J.P., 1982. Southern California landslides of
1978 and 1980. Storms, Floods, and Debris Flows in Southern
California and Arizona, 1978 and 1980. Proceedings of a Symposium,
National Research Council, and Environmental Quality
Laboratory, California Institute of Technology, Pasadena, CA,
17–18 September 1980. National Academy Press, Washington,
DC, pp. 291–319.
Soeters, R., van Westen, C.J., 1996. Slope instability, recognition,
analysis, and zonation. In: Turner, A.K., Schuster, R.L.
(Eds.), Landslides: Investigation and Mitigation, Special Report
247, Transportation Research Board, National Research
Council. National Academy Press, Washington, DC, pp. 129–
177.
Sousa, J., Voight, B., 1991. Continuum simulation of flow failures.
Geotechnique 41, 515–538.
Terlien, M.T.J., van Asch, T.W.J., van Westen, C.J., 1995. Deterministic
modelling in GIS-based landslide hazard assessment.
In: Carrara, A., Guzzetti, F. (Eds.), Geographical Information
Systems in Assessing Natural Hazards. Kluwer Academic Publishing,
The Netherlands, pp. 57–77.
Tika, T.E., Hutchinson, J.N., 1999. Ring shear tests on soil from the
Vaiont landslide slip surface. Geotechnique 49 (1), 59–74.
Tika, T.E., Vaughan, P.R., Lemos, L.J.L.J., 1996. Fast shearing of
pre-existing shear zones in soil. Geotechnique 46 (2), 197–233.
van Westen, C.J., Rengers, N., Terlien, M.T.J., Soeters, R., 1997.
Prediction of the occurrence of slope instability phenomena
through GIS-based hazard zonation. Geologische Rundschau
86, 404–414.
Wieczorek, G.F., 1984. Preparing a detailed landslide-inventory
map for hazard evaluation and reduction. Bulletin of the Association
of Engineering Geologists 21 (3), 337–342.
Wieczorek, G.F., 1987. Effect of rainfall intensity and duration on
debris flows in central Santa Cruz Mountains, California. In:
Costa, J.E., Wieczorek, G.F. (Eds.), Debris flows/Avalanches:
Process, Recognition and Mitigation. Reviews in Engineering
Geology, vol. VII. Geological Society of America, Boulder, CO,
pp. 93–104.
Wilson, R.C., Wieczorek, G.F., 1995. Rainfall threshold for the
initiation of debris flows at La Honda, California. Environmental
and Engineering Geoscience 1 (1), 11–27.
Wong, C.K.L., 1998. The New Priority Classification Systems for
Slopes and Retaining Walls. GEO Report No. 68. Geotechnical
Engineering Office, Civil Engineering Department, The Government
of the Hong Kong Special Administrative Region,
Hong Kong Government Press, Hong Kong, 117 pp.
Wong, H.N., Ho, K.K.S., Chan, Y.C., 1997. Assessment of consequence
of landslides. In: Cruden, R., Fell, R. (Eds.), Landslide
Risk Assessment. Balkema, Rotterdam, pp. 111–149.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–8786
Wright, R.H., Campbell, R.H., Nilsen, T.H., 1974. Preparation and
use of isopleth maps of landslide deposits. Geology 2, 483–485.
Wu,W.,Sidle,R.C.,1995.Adistributedslopestabilitymodelforsteep
forested basins. Water Resources Research 31 (8), 2097–2110.
Wu, T.H., Tang, W.H., Einstein, H.H., 1996. Landslide hazard and
risk assessment. In: Turner, A.K., Schuster, R.L. (Eds.), Landslides:
Investigation and Mitigation, Special Report 247, Transportation
Research Board, National Research Council. National
Academy Press, Washington, DC, pp. 106–120.
Yin, K.L., Yan, T.Z., 1988. Statistical prediction models for slope
instability of metamorphosed rocks. In: Bonnard, C. (Ed.), Proceedings
of 5th International Symposium on Landslides, Lausanne,
Switzerland, vol. 2. Balkema, Rotterdam, pp. 1269–
1272.
F.C. Dai et al. / Engineering Geology 64 (2002) 65–87 87